Generalized Autoregressive Score Models in R: The GAS Package
Responsable du projet Leopoldo Catania
Partenaire David Ardia
Kris Boudt
Résumé The GAS package for R aims to create an integrated computational environment to deal with Generalised Autoregressive Score (GAS) models. GAS models are particularly suited for univariate and multivariate time series and can be used to predict the evolution of several quantity of interest in economics and finance. Several research papers have demonstrated the superior ability of GAS models compared to alternative time series models. Most applications are in the fields of volatility modeling, systemic risk measurement, macroeconometrics, credit risk analysis and dependence modelling. The GAS package will offer the possibility to practitioners and academics to estimate GAS models using their data, perform predictions and simulations and obtain a graphical representation of the results. The data available at www.gasmodel.com will also be included for comparative and education purposes. The code will be written principally in C++ for computational purposes and then linked with R using the Rcpp, RcppArmadillo and RcppGSL packages. Code parallelisation will be included using the R parallel package and the OpenMP API. The package documentation will be written using the R roxygen2 package.
Mots-clés GAS, time series models, score models, dynamic conditional score, R software.
Page internet http://https://summerofcode.withgoogle.com/projects/#4717...
Type de projet Recherche appliquée
Domaine de recherche Financial econometrics, risk management
Source de financement Google Summer of Code
Etat Terminé
Début de projet 5-2016
Fin du projet 8-2016
Budget alloué 5,500 USD
Contact David Ardia